[Cancers detected at over-reading: analysis of data from the Loire-Atlantic breast cancer screening program, 2003-2005 (update)].

Systematic over-reading of screening mammograms increases the number of cancers detected. Published results will vary based on the particularities of different programs. A feature of the revised French program is that over-reading is limited to ACR BI-RADS 1 and 2 mammograms. In the Loire-Atlantic region, 8.7% of detected cancers between 2003-2005 were detected at over-reading. We have reviewed our database to analyze the features of these cancers detected only by the second reviewer during this period and search for causes of non-detection by the initial reviewer. Cancers detected only by the second reviewer were generally small. Breast parenchyma was mainly type 1 or 2 of the BI-RADS classification. In our study, 60% of these cancers initially missed were considered as "detection errors" because there were not detected by the initial reviewer whereas 40% were considered as "interpretation errors" because they were detected by the initial reviewer but interpreted as benign.